4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2011, 2013, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
37 #ifndef LLITE_INTERNAL_H
38 #define LLITE_INTERNAL_H
39 #include <lustre_debug.h>
40 #include <lustre_ver.h>
41 #include <lustre_disk.h> /* for s2sbi */
42 #include <lustre_eacl.h>
44 /* for struct cl_lock_descr and struct cl_io */
45 #include <cl_object.h>
47 #include <lustre_lmv.h>
48 #include <lustre_mdc.h>
49 #include <linux/lustre_intent.h>
50 #include <linux/compat.h>
56 #ifndef VM_FAULT_RETRY
57 #define VM_FAULT_RETRY 0
60 /* Kernel 3.1 kills LOOKUP_CONTINUE, LOOKUP_PARENT is equivalent to it.
61 * seem kernel commit 49084c3bb2055c401f3493c13edae14d49128ca0 */
62 #ifndef LOOKUP_CONTINUE
63 #define LOOKUP_CONTINUE LOOKUP_PARENT
66 /** Only used on client-side for indicating the tail of dir hash/offset. */
67 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
68 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
70 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
71 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
73 struct ll_dentry_data {
76 struct obd_client_handle lld_cwd_och;
77 struct obd_client_handle lld_mnt_och;
78 struct lookup_intent *lld_it;
79 unsigned int lld_sa_generation;
80 unsigned int lld_invalid:1;
81 struct rcu_head lld_rcu_head;
84 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
86 extern struct file_operations ll_pgcache_seq_fops;
88 #define LLI_INODE_MAGIC 0x111d0de5
89 #define LLI_INODE_DEAD 0xdeadd00d
91 /* remote client permission cache */
92 #define REMOTE_PERM_HASHSIZE 16
94 struct ll_getname_data {
95 #ifdef HAVE_DIR_CONTEXT
96 struct dir_context ctx;
98 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
99 struct lu_fid lgd_fid; /* target fid we are looking for */
100 int lgd_found; /* inode matched? */
103 /* llite setxid/access permission for user on remote client */
104 struct ll_remote_perm {
105 cfs_hlist_node_t lrp_list;
110 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
111 is access permission with
112 lrp_fsuid/lrp_fsgid. */
116 /* MDS has an authority for the Size-on-MDS attributes. */
117 LLIF_MDS_SIZE_LOCK = (1 << 0),
118 /* Epoch close is postponed. */
119 LLIF_EPOCH_PENDING = (1 << 1),
120 /* DONE WRITING is allowed. */
121 LLIF_DONE_WRITING = (1 << 2),
122 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
124 LLIF_SOM_DIRTY = (1 << 3),
125 /* File is contented */
126 LLIF_CONTENDED = (1 << 4),
127 /* Truncate uses server lock for this file */
128 LLIF_SRVLOCK = (1 << 5),
129 /* File data is modified. */
130 LLIF_DATA_MODIFIED = (1 << 6),
131 /* File is being restored */
132 LLIF_FILE_RESTORING = (1 << 7),
133 /* Xattr cache is attached to the file */
134 LLIF_XATTR_CACHE = (1 << 8),
137 struct ll_inode_info {
138 __u32 lli_inode_magic;
143 struct posix_acl *lli_posix_acl;
145 cfs_hlist_head_t *lli_remote_perms;
146 struct mutex lli_rmtperm_mutex;
148 /* identifying fields for both metadata and data stacks. */
149 struct lu_fid lli_fid;
150 /* Parent fid for accessing default stripe data on parent directory
151 * for allocating OST objects after a mknod() and later open-by-FID. */
152 struct lu_fid lli_pfid;
154 cfs_list_t lli_close_list;
155 cfs_list_t lli_oss_capas;
156 /* open count currently used by capability only, indicate whether
157 * capability needs renewal */
158 cfs_atomic_t lli_open_count;
159 struct obd_capa *lli_mds_capa;
160 cfs_time_t lli_rmtperm_time;
162 /* handle is to be sent to MDS later on done_writing and setattr.
163 * Open handle data are needed for the recovery to reconstruct
164 * the inode state on the MDS. XXX: recovery is not ready yet. */
165 struct obd_client_handle *lli_pending_och;
167 /* We need all three because every inode may be opened in different
169 struct obd_client_handle *lli_mds_read_och;
170 struct obd_client_handle *lli_mds_write_och;
171 struct obd_client_handle *lli_mds_exec_och;
172 __u64 lli_open_fd_read_count;
173 __u64 lli_open_fd_write_count;
174 __u64 lli_open_fd_exec_count;
175 /* Protects access to och pointers and their usage counters */
176 struct mutex lli_och_mutex;
178 struct inode lli_vfs_inode;
180 /* the most recent timestamps obtained from mds */
181 struct ost_lvb lli_lvb;
182 spinlock_t lli_agl_lock;
184 /* Try to make the d::member and f::member are aligned. Before using
185 * these members, make clear whether it is directory or not. */
189 /* serialize normal readdir and statahead-readdir. */
190 struct mutex d_readdir_mutex;
192 /* metadata statahead */
193 /* since parent-child threads can share the same @file
194 * struct, "opendir_key" is the token when dir close for
195 * case of parent exit before child -- it is me should
196 * cleanup the dir readahead. */
198 struct ll_statahead_info *d_sai;
199 struct posix_acl *d_def_acl;
200 /* protect statahead stuff. */
201 spinlock_t d_sa_lock;
202 /* "opendir_pid" is the token when lookup/revalid
203 * -- I am the owner of dir statahead. */
205 /* directory stripe information */
206 struct lmv_stripe_md *d_lsm_md;
207 /* striped directory size */
208 loff_t d_stripe_size;
209 /* striped directory nlink */
210 __u64 d_stripe_nlink;
213 #define lli_readdir_mutex u.d.d_readdir_mutex
214 #define lli_opendir_key u.d.d_opendir_key
215 #define lli_sai u.d.d_sai
216 #define lli_def_acl u.d.d_def_acl
217 #define lli_sa_lock u.d.d_sa_lock
218 #define lli_opendir_pid u.d.d_opendir_pid
219 #define lli_lsm_md u.d.d_lsm_md
220 #define lli_stripe_dir_size u.d.d_stripe_size
221 #define lli_stripe_dir_nlink u.d.d_stripe_nlink
223 /* for non-directory */
225 struct mutex f_size_mutex;
226 char *f_symlink_name;
229 * struct rw_semaphore {
230 * signed long count; // align d.d_def_acl
231 * spinlock_t wait_lock; // align d.d_sa_lock
232 * struct list_head wait_list;
235 struct rw_semaphore f_trunc_sem;
236 struct mutex f_write_mutex;
238 struct rw_semaphore f_glimpse_sem;
239 cfs_time_t f_glimpse_time;
240 cfs_list_t f_agl_list;
243 /* for writepage() only to communicate to fsync */
246 /* volatile file criteria is based on file name, this
247 * flag is used to keep the test result, so the strcmp
252 * whenever a process try to read/write the file, the
253 * jobid of the process will be saved here, and it'll
254 * be packed into the write PRC when flush later.
256 * so the read/write statistics for jobid will not be
257 * accurate if the file is shared by different jobs.
259 char f_jobid[JOBSTATS_JOBID_SIZE];
262 #define lli_size_mutex u.f.f_size_mutex
263 #define lli_symlink_name u.f.f_symlink_name
264 #define lli_maxbytes u.f.f_maxbytes
265 #define lli_trunc_sem u.f.f_trunc_sem
266 #define lli_write_mutex u.f.f_write_mutex
267 #define lli_glimpse_sem u.f.f_glimpse_sem
268 #define lli_glimpse_time u.f.f_glimpse_time
269 #define lli_agl_list u.f.f_agl_list
270 #define lli_agl_index u.f.f_agl_index
271 #define lli_async_rc u.f.f_async_rc
272 #define lli_jobid u.f.f_jobid
273 #define lli_volatile u.f.f_volatile
277 /* XXX: For following frequent used members, although they maybe special
278 * used for non-directory object, it is some time-wasting to check
279 * whether the object is directory or not before using them. On the
280 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
281 * the "ll_inode_info" size even if moving those members into u.f.
282 * So keep them out side.
284 * In the future, if more members are added only for directory,
285 * some of the following members can be moved into u.f.
288 struct cl_object *lli_clob;
290 /* mutex to request for layout lock exclusively. */
291 struct mutex lli_layout_mutex;
292 /* Layout version, protected by lli_layout_lock */
293 __u32 lli_layout_gen;
294 spinlock_t lli_layout_lock;
296 struct rw_semaphore lli_xattrs_list_rwsem;
297 struct mutex lli_xattrs_enq_lock;
298 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list */
301 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
305 spin_lock(&lli->lli_layout_lock);
306 gen = lli->lli_layout_gen;
307 spin_unlock(&lli->lli_layout_lock);
312 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
314 spin_lock(&lli->lli_layout_lock);
315 lli->lli_layout_gen = gen;
316 spin_unlock(&lli->lli_layout_lock);
319 int ll_xattr_cache_destroy(struct inode *inode);
321 int ll_xattr_cache_get(struct inode *inode,
328 * Locking to guarantee consistency of non-atomic updates to long long i_size,
329 * consistency between file size and KMS.
331 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
334 void ll_inode_size_lock(struct inode *inode);
335 void ll_inode_size_unlock(struct inode *inode);
337 // FIXME: replace the name of this with LL_I to conform to kernel stuff
338 // static inline struct ll_inode_info *LL_I(struct inode *inode)
339 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
341 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
344 /* default to about 40meg of readahead on a given system. That much tied
345 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
346 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - PAGE_CACHE_SHIFT))
348 /* default to read-ahead full files smaller than 2MB on the second read */
349 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_CACHE_SHIFT))
354 RA_STAT_DISTANT_READPAGE,
355 RA_STAT_MISS_IN_WINDOW,
356 RA_STAT_FAILED_GRAB_PAGE,
357 RA_STAT_FAILED_MATCH,
362 RA_STAT_MAX_IN_FLIGHT,
363 RA_STAT_WRONG_GRAB_PAGE,
364 RA_STAT_FAILED_REACH_END,
369 cfs_atomic_t ra_cur_pages;
370 unsigned long ra_max_pages;
371 unsigned long ra_max_pages_per_file;
372 unsigned long ra_max_read_ahead_whole_pages;
375 /* ra_io_arg will be filled in the beginning of ll_readahead with
376 * ras_lock, then the following ll_read_ahead_pages will read RA
377 * pages according to this arg, all the items in this structure are
378 * counted by page index.
381 unsigned long ria_start; /* start offset of read-ahead*/
382 unsigned long ria_end; /* end offset of read-ahead*/
383 /* If stride read pattern is detected, ria_stoff means where
384 * stride read is started. Note: for normal read-ahead, the
385 * value here is meaningless, and also it will not be accessed*/
387 /* ria_length and ria_pages are the length and pages length in the
388 * stride I/O mode. And they will also be used to check whether
389 * it is stride I/O read-ahead in the read-ahead pages*/
390 unsigned long ria_length;
391 unsigned long ria_pages;
394 /* LL_HIST_MAX=32 causes an overflow */
395 #define LL_HIST_MAX 28
396 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
397 #define LL_PROCESS_HIST_MAX 10
398 struct per_process_info {
400 struct obd_histogram pp_r_hist;
401 struct obd_histogram pp_w_hist;
404 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
405 struct ll_rw_extents_info {
406 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
409 #define LL_OFFSET_HIST_MAX 100
410 struct ll_rw_process_info {
413 loff_t rw_range_start;
415 loff_t rw_last_file_pos;
417 size_t rw_smallest_extent;
418 size_t rw_largest_extent;
419 struct ll_file_data *rw_last_file;
422 enum stats_track_type {
423 STATS_TRACK_ALL = 0, /* track all processes */
424 STATS_TRACK_PID, /* track process with this pid */
425 STATS_TRACK_PPID, /* track processes with this ppid */
426 STATS_TRACK_GID, /* track processes with this gid */
430 /* flags for sbi->ll_flags */
431 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
432 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
433 #define LL_SBI_FLOCK 0x04
434 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
435 #define LL_SBI_ACL 0x10 /* support ACL */
436 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
437 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
438 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
439 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
440 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
441 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
442 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
443 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
444 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
445 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
446 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
447 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
448 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
449 #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */
451 #define LL_SBI_FLAGS { \
474 /* default value for ll_sb_info->contention_time */
475 #define SBI_DEFAULT_CONTENTION_SECONDS 60
476 /* default value for lockless_truncate_enable */
477 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
478 #define RCE_HASHES 32
480 struct rmtacl_ctl_entry {
482 pid_t rce_key; /* hash key */
483 int rce_ops; /* acl operation type */
486 struct rmtacl_ctl_table {
488 cfs_list_t rct_entries[RCE_HASHES];
495 pid_t ee_key; /* hash key */
496 struct lu_fid ee_fid;
497 int ee_type; /* ACL type for ACCESS or DEFAULT */
498 ext_acl_xattr_header *ee_acl;
503 cfs_list_t et_entries[EE_HASHES];
508 /* this protects pglist and ra_info. It isn't safe to
509 * grab from interrupt contexts */
511 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
512 spinlock_t ll_process_lock; /* ll_rw_process_info */
513 struct obd_uuid ll_sb_uuid;
514 struct obd_export *ll_md_exp;
515 struct obd_export *ll_dt_exp;
516 struct proc_dir_entry* ll_proc_root;
517 struct lu_fid ll_root_fid; /* root object fid */
520 unsigned int ll_umounting:1,
521 ll_xattr_cache_enabled:1;
522 cfs_list_t ll_conn_chain; /* per-conn chain of SBs */
523 struct lustre_client_ocd ll_lco;
525 cfs_list_t ll_orphan_dentry_list; /*please don't ask -p*/
526 struct ll_close_queue *ll_lcq;
528 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
530 /* Used to track "unstable" pages on a client, and maintain a
531 * LRU list of clean pages. An "unstable" page is defined as
532 * any page which is sent to a server as part of a bulk request,
533 * but is uncommitted to stable storage. */
534 struct cl_client_cache ll_cache;
536 struct lprocfs_stats *ll_ra_stats;
538 struct ll_ra_info ll_ra_info;
539 unsigned int ll_namelen;
540 struct file_operations *ll_fop;
542 /* =0 - hold lock over whole read/write
543 * >0 - max. chunk to be read/written w/o lock re-acquiring */
544 unsigned long ll_max_rw_chunk;
545 unsigned int ll_md_brw_size; /* used by readdir */
547 struct lu_site *ll_site;
548 struct cl_device *ll_cl;
550 struct ll_rw_extents_info ll_rw_extents_info;
551 int ll_extent_process_count;
552 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
553 unsigned int ll_offset_process_count;
554 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
555 unsigned int ll_rw_offset_entry_count;
556 int ll_stats_track_id;
557 enum stats_track_type ll_stats_track_type;
560 /* metadata stat-ahead */
561 unsigned int ll_sa_max; /* max statahead RPCs */
562 atomic_t ll_sa_total; /* statahead thread started
564 atomic_t ll_sa_wrong; /* statahead thread stopped for
566 atomic_t ll_agl_total; /* AGL thread started count */
568 dev_t ll_sdev_orig; /* save s_dev before assign for
570 struct rmtacl_ctl_table ll_rct;
571 struct eacl_table ll_et;
574 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
579 struct task_struct *lrr_reader;
580 cfs_list_t lrr_linkage;
584 * per file-descriptor read-ahead data.
586 struct ll_readahead_state {
589 * index of the last page that read(2) needed and that wasn't in the
590 * cache. Used by ras_update() to detect seeks.
592 * XXX nikita: if access seeks into cached region, Lustre doesn't see
595 unsigned long ras_last_readpage;
597 * number of pages read after last read-ahead window reset. As window
598 * is reset on each seek, this is effectively a number of consecutive
599 * accesses. Maybe ->ras_accessed_in_window is better name.
601 * XXX nikita: window is also reset (by ras_update()) when Lustre
602 * believes that memory pressure evicts read-ahead pages. In that
603 * case, it probably doesn't make sense to expand window to
604 * PTLRPC_MAX_BRW_PAGES on the third access.
606 unsigned long ras_consecutive_pages;
608 * number of read requests after the last read-ahead window reset
609 * As window is reset on each seek, this is effectively the number
610 * on consecutive read request and is used to trigger read-ahead.
612 unsigned long ras_consecutive_requests;
614 * Parameters of current read-ahead window. Handled by
615 * ras_update(). On the initial access to the file or after a seek,
616 * window is reset to 0. After 3 consecutive accesses, window is
617 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
618 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
620 unsigned long ras_window_start, ras_window_len;
622 * Where next read-ahead should start at. This lies within read-ahead
623 * window. Read-ahead window is read in pieces rather than at once
624 * because: 1. lustre limits total number of pages under read-ahead by
625 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
626 * not covered by DLM lock.
628 unsigned long ras_next_readahead;
630 * Total number of ll_file_read requests issued, reads originating
631 * due to mmap are not counted in this total. This value is used to
632 * trigger full file read-ahead after multiple reads to a small file.
634 unsigned long ras_requests;
636 * Page index with respect to the current request, these value
637 * will not be accurate when dealing with reads issued via mmap.
639 unsigned long ras_request_index;
641 * list of struct ll_ra_read's one per read(2) call current in
642 * progress against this file descriptor. Used by read-ahead code,
643 * protected by ->ras_lock.
645 cfs_list_t ras_read_beads;
647 * The following 3 items are used for detecting the stride I/O
649 * In stride I/O mode,
650 * ...............|-----data-----|****gap*****|--------|******|....
651 * offset |-stride_pages-|-stride_gap-|
652 * ras_stride_offset = offset;
653 * ras_stride_length = stride_pages + stride_gap;
654 * ras_stride_pages = stride_pages;
655 * Note: all these three items are counted by pages.
657 unsigned long ras_stride_length;
658 unsigned long ras_stride_pages;
659 pgoff_t ras_stride_offset;
661 * number of consecutive stride request count, and it is similar as
662 * ras_consecutive_requests, but used for stride I/O mode.
663 * Note: only more than 2 consecutive stride request are detected,
664 * stride read-ahead will be enable
666 unsigned long ras_consecutive_stride_requests;
669 extern struct kmem_cache *ll_file_data_slab;
670 struct lustre_handle;
671 struct ll_file_data {
672 struct ll_readahead_state fd_ras;
673 struct ccc_grouplock fd_grouplock;
677 /* openhandle if lease exists for this file.
678 * Borrow lli->lli_och_mutex to protect assignment */
679 struct obd_client_handle *fd_lease_och;
680 struct obd_client_handle *fd_och;
681 struct file *fd_file;
682 /* Indicate whether need to report failure when close.
683 * true: failure is known, not report again.
684 * false: unknown failure, should report. */
685 bool fd_write_failed;
688 struct lov_stripe_md;
690 extern spinlock_t inode_lock;
692 extern struct proc_dir_entry *proc_lustre_fs_root;
694 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
696 return &lli->lli_vfs_inode;
700 struct inode *icbd_parent;
701 struct dentry **icbd_childp;
705 __u32 ll_i2suppgid(struct inode *i);
706 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
708 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
710 #if BITS_PER_LONG == 32
717 (sbi->ll_flags & LL_SBI_32BIT_API)
722 #define LLAP_MAGIC 98764321
724 extern struct kmem_cache *ll_async_page_slab;
725 extern size_t ll_async_page_slab_size;
727 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
728 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
729 struct ll_ra_read *ll_ra_read_get(struct file *f);
731 /* llite/lproc_llite.c */
733 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
734 struct super_block *sb, char *osc, char *mdc);
735 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
736 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
737 extern struct lprocfs_seq_vars lprocfs_llite_obd_vars[];
739 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
740 struct super_block *sb, char *osc, char *mdc){return 0;}
741 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
742 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
747 extern struct file_operations ll_dir_operations;
748 extern struct inode_operations ll_dir_inode_operations;
749 #ifdef HAVE_DIR_CONTEXT
750 int ll_dir_read(struct inode *inode, struct md_op_data *op_data,
751 struct dir_context *ctx);
753 int ll_dir_read(struct inode *inode, struct md_op_data *op_data,
754 void *cookie, filldir_t filldir);
756 int ll_get_mdt_idx(struct inode *inode);
757 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
759 struct lu_dirent *ll_dir_entry_start(struct inode *dir,
760 struct md_op_data *op_data,
761 struct page **ppage);
763 struct lu_dirent *ll_dir_entry_next(struct inode *dir,
764 struct md_op_data *op_data,
765 struct lu_dirent *ent,
766 struct page **ppage);
768 int ll_objects_destroy(struct ptlrpc_request *request,
770 struct inode *ll_iget(struct super_block *sb, ino_t hash,
771 struct lustre_md *lic);
772 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
773 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
774 void *data, int flag);
775 #ifndef HAVE_IOP_ATOMIC_OPEN
776 struct lookup_intent *ll_convert_intent(struct open_intent *oit,
779 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
780 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
781 int ll_d_mountpoint(struct dentry *dparent, struct dentry *dchild,
783 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
786 int ll_writepage(struct page *page, struct writeback_control *wbc);
787 int ll_writepages(struct address_space *, struct writeback_control *wbc);
788 void ll_removepage(struct page *page);
789 int ll_readpage(struct file *file, struct page *page);
790 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
791 int ll_file_punch(struct inode *, loff_t, int);
792 ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int);
793 void ll_clear_file_contended(struct inode*);
794 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
795 int ll_readahead(const struct lu_env *env, struct cl_io *io,
796 struct cl_page_list *queue, struct ll_readahead_state *ras,
798 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
799 struct ll_cl_context *ll_cl_init(struct file *file, struct page *vmpage);
800 void ll_cl_fini(struct ll_cl_context *lcc);
803 extern struct file_operations ll_file_operations;
804 extern struct file_operations ll_file_operations_flock;
805 extern struct file_operations ll_file_operations_noflock;
806 extern struct inode_operations ll_file_inode_operations;
807 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
809 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
810 ldlm_mode_t l_req_mode);
811 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
812 struct lustre_handle *lockh, __u64 flags,
814 int __ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
816 #ifdef HAVE_IOP_ATOMIC_OPEN
817 int ll_revalidate_nd(struct dentry *dentry, unsigned int flags);
819 int ll_revalidate_nd(struct dentry *dentry, struct nameidata *nd);
821 int ll_file_open(struct inode *inode, struct file *file);
822 int ll_file_release(struct inode *inode, struct file *file);
823 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
824 struct lov_stripe_md *lsm, lstat_t *st);
825 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
826 int ll_local_open(struct file *file,
827 struct lookup_intent *it, struct ll_file_data *fd,
828 struct obd_client_handle *och);
829 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
830 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
832 int ll_md_real_close(struct inode *inode, fmode_t fmode);
833 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
834 struct obd_client_handle **och, unsigned long flags);
835 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
836 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
837 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
838 __u64 ioepoch, int sync);
839 int ll_md_setattr(struct dentry *dentry, struct md_op_data *op_data,
840 struct md_open_data **mod);
841 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
842 struct lustre_handle *fh);
843 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
844 struct ll_file_data *file, loff_t pos,
845 size_t count, int rw);
846 int ll_getattr_it(struct vfsmount *mnt, struct dentry *de,
847 struct lookup_intent *it, struct kstat *stat);
848 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
849 struct ll_file_data *ll_file_data_get(void);
850 struct posix_acl * ll_get_acl(struct inode *inode, int type);
851 int ll_migrate(struct inode *parent, struct file *file, int mdtidx,
852 const char *name, int namelen);
853 int ll_get_fid_by_name(struct inode *parent, const char *name,
854 int namelen, struct lu_fid *fid);
855 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
856 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
858 # ifndef HAVE_INODE_PERMISION_2ARGS
859 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
861 int ll_inode_permission(struct inode *inode, int mask);
865 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
866 __u64 flags, struct lov_user_md *lum,
868 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
869 struct lov_mds_md **lmm, int *lmm_size,
870 struct ptlrpc_request **request);
871 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
873 int ll_dir_getstripe(struct inode *inode, void **lmmp,
874 int *lmm_size, struct ptlrpc_request **request,
876 #ifdef HAVE_FILE_FSYNC_4ARGS
877 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
878 #elif defined(HAVE_FILE_FSYNC_2ARGS)
879 int ll_fsync(struct file *file, int data);
881 int ll_fsync(struct file *file, struct dentry *dentry, int data);
883 int ll_do_fiemap(struct inode *inode, struct ll_user_fiemap *fiemap,
885 int ll_merge_lvb(const struct lu_env *env, struct inode *inode);
886 int ll_get_grouplock(struct inode *inode, struct file *file, unsigned long arg);
887 int ll_put_grouplock(struct inode *inode, struct file *file, unsigned long arg);
888 int ll_fid2path(struct inode *inode, void *arg);
889 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
890 int ll_hsm_release(struct inode *inode);
892 struct obd_client_handle *ll_lease_open(struct inode *inode, struct file *file,
893 fmode_t mode, __u64 flags);
894 int ll_lease_close(struct obd_client_handle *och, struct inode *inode,
899 int ll_d_init(struct dentry *de);
900 extern struct dentry_operations ll_d_ops;
901 void ll_intent_drop_lock(struct lookup_intent *);
902 void ll_intent_release(struct lookup_intent *);
903 void ll_invalidate_aliases(struct inode *);
904 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
905 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
906 int ll_revalidate_it_finish(struct ptlrpc_request *request,
907 struct lookup_intent *it, struct dentry *de);
909 /* llite/llite_lib.c */
910 extern struct super_operations lustre_super_operations;
912 char *ll_read_opt(const char *opt, char *data);
913 void ll_lli_init(struct ll_inode_info *lli);
914 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
915 void ll_put_super(struct super_block *sb);
916 void ll_kill_super(struct super_block *sb);
917 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
918 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
919 void ll_clear_inode(struct inode *inode);
920 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr, bool hsm_import);
921 int ll_setattr(struct dentry *de, struct iattr *attr);
922 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
923 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
924 __u64 max_age, __u32 flags);
925 void ll_update_inode(struct inode *inode, struct lustre_md *md);
926 void ll_read_inode2(struct inode *inode, void *opaque);
927 void ll_delete_inode(struct inode *inode);
928 int ll_iocontrol(struct inode *inode, struct file *file,
929 unsigned int cmd, unsigned long arg);
930 int ll_flush_ctx(struct inode *inode);
931 void ll_umount_begin(struct super_block *sb);
932 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
933 #ifdef HAVE_SUPEROPS_USE_DENTRY
934 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
936 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
938 void ll_dirty_page_discard_warn(struct page *page, int ioret);
939 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
940 struct super_block *, struct lookup_intent *);
941 void lustre_dump_dentry(struct dentry *, int recur);
942 void lustre_dump_inode(struct inode *);
943 int ll_obd_statfs(struct inode *inode, void *arg);
944 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
945 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
946 int ll_get_max_cookiesize(struct ll_sb_info *sbi, int *max_cookiesize);
947 int ll_get_default_cookiesize(struct ll_sb_info *sbi, int *default_cookiesize);
948 int ll_process_config(struct lustre_cfg *lcfg);
949 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
950 struct inode *i1, struct inode *i2,
951 const char *name, int namelen,
952 int mode, __u32 opc, void *data);
953 void ll_finish_md_op_data(struct md_op_data *op_data);
954 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
955 char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
957 /* llite/llite_nfs.c */
958 extern struct export_operations lustre_export_operations;
959 __u32 get_uuid2int(const char *name, int len);
960 struct inode *search_inode_for_lustre(struct super_block *sb,
961 const struct lu_fid *fid);
963 /* llite/special.c */
964 extern struct inode_operations ll_special_inode_operations;
965 extern struct file_operations ll_special_chr_inode_fops;
966 extern struct file_operations ll_special_chr_file_fops;
967 extern struct file_operations ll_special_blk_inode_fops;
968 extern struct file_operations ll_special_fifo_inode_fops;
969 extern struct file_operations ll_special_fifo_file_fops;
970 extern struct file_operations ll_special_sock_inode_fops;
972 /* llite/symlink.c */
973 extern struct inode_operations ll_fast_symlink_inode_operations;
975 /* llite/llite_close.c */
976 struct ll_close_queue {
979 wait_queue_head_t lcq_waitq;
980 struct completion lcq_comp;
981 cfs_atomic_t lcq_stop;
984 struct ccc_object *cl_inode2ccc(struct inode *inode);
987 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
988 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
990 /* specific achitecture can implement only part of this list */
991 enum vvp_io_subtype {
994 /** io started from splice_{read|write} */
1001 enum vvp_io_subtype cui_io_subtype;
1005 struct pipe_inode_info *cui_pipe;
1006 unsigned int cui_flags;
1008 struct vvp_fault_io {
1010 * Inode modification time that is checked across DLM
1014 struct vm_area_struct *ft_vma;
1016 * locked page returned from vvp_io
1018 struct page *ft_vmpage;
1019 struct vm_fault_api {
1023 struct vm_fault *ft_vmf;
1025 * fault API used bitflags for return code.
1027 unsigned int ft_flags;
1032 * Read-ahead state used by read and page-fault IO contexts.
1034 struct ll_ra_read cui_bead;
1036 * Set when cui_bead has been initialized.
1038 int cui_ra_window_set;
1040 * Partially truncated page, that vvp_io_trunc_start() keeps locked
1043 struct cl_page *cui_partpage;
1047 * IO arguments for various VFS I/O interfaces.
1049 struct vvp_io_args {
1050 /** normal/sendfile/splice */
1051 enum vvp_io_subtype via_io_subtype;
1055 struct kiocb *via_iocb;
1056 struct iovec *via_iov;
1057 unsigned long via_nrsegs;
1060 struct pipe_inode_info *via_pipe;
1061 unsigned int via_flags;
1066 struct ll_cl_context {
1068 struct cl_io *lcc_io;
1069 struct cl_page *lcc_page;
1070 struct lu_env *lcc_env;
1075 struct vvp_thread_info {
1076 struct ost_lvb vti_lvb;
1077 struct cl_2queue vti_queue;
1078 struct iovec vti_local_iov;
1079 struct vvp_io_args vti_args;
1080 struct ra_io_arg vti_ria;
1081 struct kiocb vti_kiocb;
1082 struct ll_cl_context vti_io_ctx;
1085 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
1087 extern struct lu_context_key vvp_key;
1088 struct vvp_thread_info *info;
1090 info = lu_context_key_get(&env->le_ctx, &vvp_key);
1091 LASSERT(info != NULL);
1095 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
1096 enum vvp_io_subtype type)
1098 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
1100 ret->via_io_subtype = type;
1105 struct vvp_session {
1106 struct vvp_io vs_ios;
1109 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
1111 extern struct lu_context_key vvp_session_key;
1112 struct vvp_session *ses;
1114 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
1115 LASSERT(ses != NULL);
1119 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
1121 return &vvp_env_session(env)->vs_ios;
1124 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
1125 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
1126 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
1128 /* llite/llite_mmap.c */
1129 typedef struct rb_root rb_root_t;
1130 typedef struct rb_node rb_node_t;
1132 struct ll_lock_tree_node;
1133 struct ll_lock_tree {
1135 cfs_list_t lt_locked_list;
1136 struct ll_file_data *lt_fd;
1139 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1140 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1141 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
1142 __u64 end, ldlm_mode_t mode);
1143 void policy_from_vma(ldlm_policy_data_t *policy,
1144 struct vm_area_struct *vma, unsigned long addr, size_t count);
1145 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1148 static inline void ll_invalidate_page(struct page *vmpage)
1150 struct address_space *mapping = vmpage->mapping;
1151 loff_t offset = vmpage->index << PAGE_CACHE_SHIFT;
1153 LASSERT(PageLocked(vmpage));
1154 if (mapping == NULL)
1158 * truncate_complete_page() calls
1159 * a_ops->invalidatepage()->cl_page_delete()->vvp_page_delete().
1161 ll_teardown_mmaps(mapping, offset, offset + PAGE_CACHE_SIZE);
1162 truncate_complete_page(mapping, vmpage);
1165 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1167 /* don't need an addref as the sb_info should be holding one */
1168 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1170 return ll_s2sbi(sb)->ll_dt_exp;
1173 /* don't need an addref as the sb_info should be holding one */
1174 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1176 return ll_s2sbi(sb)->ll_md_exp;
1179 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1181 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1187 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1188 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1190 return ll_s2sbi(inode->i_sb);
1193 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1195 return ll_s2dtexp(inode->i_sb);
1198 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1200 return ll_s2mdexp(inode->i_sb);
1203 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1207 LASSERT(inode != NULL);
1208 fid = &ll_i2info(inode)->lli_fid;
1213 static inline __u64 ll_file_maxbytes(struct inode *inode)
1215 return ll_i2info(inode)->lli_maxbytes;
1219 int ll_setxattr(struct dentry *dentry, const char *name,
1220 const void *value, size_t size, int flags);
1221 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1222 void *buffer, size_t size);
1223 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1224 int ll_removexattr(struct dentry *dentry, const char *name);
1226 /* llite/remote_perm.c */
1227 extern struct kmem_cache *ll_remote_perm_cachep;
1228 extern struct kmem_cache *ll_rmtperm_hash_cachep;
1230 cfs_hlist_head_t *alloc_rmtperm_hash(void);
1231 void free_rmtperm_hash(cfs_hlist_head_t *hash);
1232 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1233 int lustre_check_remote_perm(struct inode *inode, int mask);
1235 /* llite/llite_capa.c */
1236 extern struct timer_list ll_capa_timer;
1238 int ll_capa_thread_start(void);
1239 void ll_capa_thread_stop(void);
1240 void ll_capa_timer_callback(unsigned long unused);
1242 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1243 int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa);
1245 void ll_capa_open(struct inode *inode);
1246 void ll_capa_close(struct inode *inode);
1248 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1249 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1251 void ll_truncate_free_capa(struct obd_capa *ocapa);
1252 void ll_clear_inode_capas(struct inode *inode);
1253 void ll_print_capa_stat(struct ll_sb_info *sbi);
1255 /* llite/llite_cl.c */
1256 extern struct lu_device_type vvp_device_type;
1259 * Common IO arguments for various VFS I/O interfaces.
1261 int cl_sb_init(struct super_block *sb);
1262 int cl_sb_fini(struct super_block *sb);
1263 enum cl_lock_mode vvp_mode_from_vma(struct vm_area_struct *vma);
1264 void ll_io_init(struct cl_io *io, const struct file *file, int write);
1266 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1267 struct ll_readahead_state *ras, unsigned long index,
1269 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1270 int ll_is_file_contended(struct file *file);
1271 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1273 /* llite/llite_rmtacl.c */
1274 #ifdef CONFIG_FS_POSIX_ACL
1275 obd_valid rce_ops2valid(int ops);
1276 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1277 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1278 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1279 void rct_init(struct rmtacl_ctl_table *rct);
1280 void rct_fini(struct rmtacl_ctl_table *rct);
1282 void ee_free(struct eacl_entry *ee);
1283 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1284 ext_acl_xattr_header *header);
1285 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1286 struct lu_fid *fid, int type);
1287 void et_search_free(struct eacl_table *et, pid_t key);
1288 void et_init(struct eacl_table *et);
1289 void et_fini(struct eacl_table *et);
1294 #define LL_SA_RPC_MIN 2
1295 #define LL_SA_RPC_DEF 32
1296 #define LL_SA_RPC_MAX 8192
1298 #define LL_SA_CACHE_BIT 5
1299 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1300 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1302 /* per inode struct, for dir only */
1303 struct ll_statahead_info {
1304 struct inode *sai_inode;
1305 cfs_atomic_t sai_refcount; /* when access this struct, hold
1307 unsigned int sai_generation; /* generation for statahead */
1308 unsigned int sai_max; /* max ahead of lookup */
1309 __u64 sai_sent; /* stat requests sent count */
1310 __u64 sai_replied; /* stat requests which received
1312 __u64 sai_index; /* index of statahead entry */
1313 __u64 sai_index_wait; /* index of entry which is the
1314 * caller is waiting for */
1315 __u64 sai_hit; /* hit count */
1316 __u64 sai_miss; /* miss count:
1317 * for "ls -al" case, it includes
1318 * hidden dentry miss;
1319 * for "ls -l" case, it does not
1320 * include hidden dentry miss.
1321 * "sai_miss_hidden" is used for
1324 unsigned int sai_consecutive_miss; /* consecutive miss */
1325 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1326 * is not a hidden one */
1327 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1328 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1330 sai_in_readpage:1,/* statahead is in readdir()*/
1331 sai_agl_valid:1;/* AGL is valid for the dir */
1332 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1333 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1334 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1335 cfs_list_t sai_entries; /* entry list */
1336 cfs_list_t sai_entries_received; /* entries returned */
1337 cfs_list_t sai_entries_stated; /* entries stated */
1338 cfs_list_t sai_entries_agl; /* AGL entries to be sent */
1339 cfs_list_t sai_cache[LL_SA_CACHE_SIZE];
1340 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1341 cfs_atomic_t sai_cache_count; /* entry count in cache */
1344 int do_statahead_enter(struct inode *dir, struct dentry **dentry,
1346 void ll_stop_statahead(struct inode *dir, void *key);
1348 static inline int ll_glimpse_size(struct inode *inode)
1350 struct ll_inode_info *lli = ll_i2info(inode);
1353 down_read(&lli->lli_glimpse_sem);
1354 rc = cl_glimpse_size(inode);
1355 lli->lli_glimpse_time = cfs_time_current();
1356 up_read(&lli->lli_glimpse_sem);
1361 ll_statahead_mark(struct inode *dir, struct dentry *dentry)
1363 struct ll_inode_info *lli = ll_i2info(dir);
1364 struct ll_statahead_info *sai = lli->lli_sai;
1365 struct ll_dentry_data *ldd = ll_d2d(dentry);
1367 /* not the same process, don't mark */
1368 if (lli->lli_opendir_pid != current_pid())
1371 LASSERT(ldd != NULL);
1373 ldd->lld_sa_generation = sai->sai_generation;
1377 d_need_statahead(struct inode *dir, struct dentry *dentryp)
1379 struct ll_inode_info *lli;
1380 struct ll_dentry_data *ldd;
1382 if (ll_i2sbi(dir)->ll_sa_max == 0)
1385 lli = ll_i2info(dir);
1386 /* not the same process, don't statahead */
1387 if (lli->lli_opendir_pid != current_pid())
1390 /* statahead has been stopped */
1391 if (lli->lli_opendir_key == NULL)
1394 ldd = ll_d2d(dentryp);
1396 * When stats a dentry, the system trigger more than once "revalidate"
1397 * or "lookup", for "getattr", for "getxattr", and maybe for others.
1398 * Under patchless client mode, the operation intent is not accurate,
1399 * which maybe misguide the statahead thread. For example:
1400 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1401 * have the same operation intent -- "IT_GETATTR".
1402 * In fact, one dentry should has only one chance to interact with the
1403 * statahead thread, otherwise the statahead windows will be confused.
1404 * The solution is as following:
1405 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1406 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1407 * will bypass interacting with statahead thread for checking:
1408 * "lld_sa_generation == lli_sai->sai_generation"
1410 if (ldd && lli->lli_sai &&
1411 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1418 ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int only_unplug)
1422 ret = d_need_statahead(dir, *dentryp);
1426 return do_statahead_enter(dir, dentryp, only_unplug);
1429 /* llite ioctl register support rountine */
1436 #define LLIOC_MAX_CMD 256
1439 * Rules to write a callback function:
1442 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1443 * returned to ll_iocontrol_register. Callback functions should use this
1444 * data to check the potential collasion of ioctl cmd. If collasion is
1445 * found, callback function should return LLIOC_CONT.
1446 * @rcp: The result of ioctl command.
1449 * If @magic matches the pointer returned by ll_iocontrol_data, the
1450 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1452 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1453 struct file *file, unsigned int cmd, unsigned long arg,
1454 void *magic, int *rcp);
1456 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
1457 unsigned int cmd, unsigned long arg, int *rcp);
1459 /* export functions */
1460 /* Register ioctl block dynamatically for a regular file.
1462 * @cmd: the array of ioctl command set
1463 * @count: number of commands in the @cmd
1464 * @cb: callback function, it will be called if an ioctl command is found to
1465 * belong to the command list @cmd.
1468 * A magic pointer will be returned if success;
1469 * otherwise, NULL will be returned.
1471 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1472 void ll_iocontrol_unregister(void *magic);
1476 /* lclient compat stuff */
1477 #define cl_inode_info ll_inode_info
1478 #define cl_i2info(info) ll_i2info(info)
1479 #define cl_inode_mode(inode) ((inode)->i_mode)
1480 #define cl_i2sbi ll_i2sbi
1482 static inline struct ll_file_data *cl_iattr2fd(struct inode *inode,
1483 const struct iattr *attr)
1485 LASSERT(attr->ia_valid & ATTR_FILE);
1486 return LUSTRE_FPRIVATE(attr->ia_file);
1489 static inline void cl_isize_lock(struct inode *inode)
1491 ll_inode_size_lock(inode);
1494 static inline void cl_isize_unlock(struct inode *inode)
1496 ll_inode_size_unlock(inode);
1499 static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms)
1501 LASSERT(mutex_is_locked(&ll_i2info(inode)->lli_size_mutex));
1502 i_size_write(inode, kms);
1505 static inline void cl_isize_write(struct inode *inode, loff_t kms)
1507 ll_inode_size_lock(inode);
1508 i_size_write(inode, kms);
1509 ll_inode_size_unlock(inode);
1512 #define cl_isize_read(inode) i_size_read(inode)
1514 static inline int cl_merge_lvb(const struct lu_env *env, struct inode *inode)
1516 return ll_merge_lvb(env, inode);
1519 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1520 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1521 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1523 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1525 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1526 enum cl_fsync_mode mode, int ignore_layout);
1528 /** direct write pages */
1529 struct ll_dio_pages {
1530 /** page array to be written. we don't support
1531 * partial pages except the last one. */
1532 struct page **ldp_pages;
1533 /* offset of each page */
1534 loff_t *ldp_offsets;
1535 /** if ldp_offsets is NULL, it means a sequential
1536 * pages to be written, then this is the file offset
1537 * of the * first page. */
1538 loff_t ldp_start_offset;
1539 /** how many bytes are to be written. */
1541 /** # of pages in the array. */
1545 static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt,
1548 int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ :
1551 ll_stats_ops_tally(ll_s2sbi(cl2ccc_dev(dev)->cdv_sb), opc, rc);
1554 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1555 int rw, struct inode *inode,
1556 struct ll_dio_pages *pv);
1558 static inline int ll_file_nolock(const struct file *file)
1560 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1561 struct inode *inode = file->f_dentry->d_inode;
1563 LASSERT(fd != NULL);
1564 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1565 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1568 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1569 struct lookup_intent *it, __u64 *bits)
1571 if (!it->d.lustre.it_lock_set) {
1572 struct lustre_handle handle;
1574 /* If this inode is a remote object, it will get two
1575 * separate locks in different namespaces, Master MDT,
1576 * where the name entry is, will grant LOOKUP lock,
1577 * remote MDT, where the object is, will grant
1578 * UPDATE|PERM lock. The inode will be attched to both
1579 * LOOKUP and PERM locks, so revoking either locks will
1580 * case the dcache being cleared */
1581 if (it->d.lustre.it_remote_lock_mode) {
1582 handle.cookie = it->d.lustre.it_remote_lock_handle;
1583 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1584 "(%p) for remote lock "LPX64"\n",
1585 PFID(ll_inode2fid(inode)), inode,
1587 md_set_lock_data(exp, &handle.cookie, inode, NULL);
1590 handle.cookie = it->d.lustre.it_lock_handle;
1592 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"(%p)"
1593 " for lock "LPX64"\n",
1594 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1596 md_set_lock_data(exp, &handle.cookie, inode,
1597 &it->d.lustre.it_lock_bits);
1598 it->d.lustre.it_lock_set = 1;
1602 *bits = it->d.lustre.it_lock_bits;
1605 static inline void ll_lock_dcache(struct inode *inode)
1607 #ifdef HAVE_DCACHE_LOCK
1608 spin_lock(&dcache_lock);
1610 spin_lock(&inode->i_lock);
1614 static inline void ll_unlock_dcache(struct inode *inode)
1616 #ifdef HAVE_DCACHE_LOCK
1617 spin_unlock(&dcache_lock);
1619 spin_unlock(&inode->i_lock);
1623 static inline int d_lustre_invalid(const struct dentry *dentry)
1625 struct ll_dentry_data *lld = ll_d2d(dentry);
1627 return (lld == NULL) || lld->lld_invalid;
1630 static inline void __d_lustre_invalidate(struct dentry *dentry)
1632 struct ll_dentry_data *lld = ll_d2d(dentry);
1635 lld->lld_invalid = 1;
1639 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1640 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1641 * else dput() of the last refcount will unhash this dentry and kill it.
1643 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1645 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1646 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1647 dentry->d_parent, dentry->d_inode, d_count(dentry));
1649 spin_lock_nested(&dentry->d_lock,
1650 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1651 __d_lustre_invalidate(dentry);
1652 if (d_count(dentry) == 0)
1654 spin_unlock(&dentry->d_lock);
1657 static inline void d_lustre_revalidate(struct dentry *dentry)
1659 spin_lock(&dentry->d_lock);
1660 LASSERT(ll_d2d(dentry) != NULL);
1661 ll_d2d(dentry)->lld_invalid = 0;
1662 spin_unlock(&dentry->d_lock);
1665 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0)
1666 /* Compatibility for old (1.8) compiled userspace quota code */
1667 struct if_quotactl_18 {
1672 struct obd_dqinfo qc_dqinfo;
1673 struct obd_dqblk qc_dqblk;
1675 struct obd_uuid obd_uuid;
1677 #define LL_IOC_QUOTACTL_18 _IOWR('f', 162, struct if_quotactl_18 *)
1678 /* End compatibility for old (1.8) compiled userspace quota code */
1680 #warning "remove old LL_IOC_QUOTACTL_18 compatibility code"
1681 #endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0) */
1684 LL_LAYOUT_GEN_NONE = ((__u32)-2), /* layout lock was cancelled */
1685 LL_LAYOUT_GEN_EMPTY = ((__u32)-1) /* for empty layout */
1688 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1689 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1690 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1692 int ll_xattr_init(void);
1693 void ll_xattr_fini(void);
1695 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1696 struct cl_page *page, enum cl_req_type crt);
1698 #endif /* LLITE_INTERNAL_H */